Glossary · 32 terms

Editorial intelligence glossary

Short, declarative definitions of the 32 core terms used across EarlyForge — editorial intelligence, real-time trend detection, generative SEO and the surrounding workflow. Each entry links to the related concepts so the graph stays navigable.

Editorial intelligence — the field

Editorial intelligence#: The discipline of applying machine learning, real-time data pipelines and editorial judgment to detect, qualify and cover emerging news, cultural and commercial trends. Where business intelligence answers 'what happened in our company', editorial intelligence answers 'what is happening in the world that we should publish about'.
Real-time trend detection#: The continuous observation of news, social, editorial and structured-data sources to identify emerging topics within minutes of their formation, rather than after they appear in search-volume tools. Real-time detection is the upstream alternative to retrospective discovery.
Generative SEO (GEO)#: Search engine optimization aimed at large language models — ChatGPT, Claude, Perplexity, Gemini, Copilot — rather than at traditional SERPs. GEO tactics include structured data, llms.txt, brand entity disambiguation via Wikidata, and answer-first content formats designed for extractive citation.
Newsroom software#: Production tooling used by editorial teams to plan, draft, edit and publish stories. Modern newsroom software increasingly bundles trend detection, brief generation and multi-CMS publishing into one workflow.
Answer engine optimization (AEO)#: A subset of GEO focused on making content extractable by answer engines (Perplexity, ChatGPT browse, Google AI Overviews). Tactics include FAQ schema, definitional short-form paragraphs, citation-ready statistics, and explicit entity declarations.

Signals & sources

Upstream signal layer#: The set of editorial and social channels — news wires, regional press, journalistic Twitter/X, niche subreddits, podcast transcripts, scientific preprints — where trends first surface, hours or days before they generate measurable search volume.
Downstream signal layer#: Aggregated search-query and shopping-intent data (search interest, marketplace queries) that reflects what consumers are actively researching. Complementary to the upstream signal layer, which observes the conversations and coverage that precede search behaviour.
Connector#: A software adapter that ingests a single external data source into a trend-detection pipeline. EarlyForge monitors thousands of sources across news APIs, social platforms, RSS feeds, scientific repositories and regional press networks.
Data source#: Any external service or repository that produces structured or semi-structured signals consumable by a connector. Examples include the Associated Press feed, Reddit, arXiv, and country-specific news aggregators.
Source trust score#: A numeric weight applied per data source to balance signal quantity against editorial reliability. A primary news wire receives a higher trust score than an unverified social account.

Scoring & ranking

Velocity score#: The rate at which mentions of a topic are accelerating across observed sources, measured in derivative terms (mentions per unit time per unit time). High velocity early in a topic's life suggests a trend on the rise.
Diversity score#: A measure of how many independent sources, regions and languages are mentioning the same underlying event. Wide diversity distinguishes a global story from a local one, and a real trend from coordinated amplification.
Cohesion score#: A measure of whether the mentions of a topic genuinely talk about the same story across sources and languages. Used to demote noisy or fragmented signals and amplify coherent, well-anchored stories.
Global score#: The composite score that ranks trends on Pulse and inside EarlyForge. Combines velocity, diversity, recency, cohesion and SEO opportunity into a single 0–100 number.

Aggregation & clustering

Root event#: The underlying real-world event that multiple parallel mentions are all describing — across languages, regions and source types. Root-event clustering merges these mentions into a single trackable cluster, so coverage decisions are made on the whole story instead of one language silo.
Trend cluster#: A group of signals that have been algorithmically determined to refer to the same root event. Each cluster carries a global score, a representative headline, and a list of contributing sources.
Cross-lingual merge#: The deduplication step that recognises the same root event across 28+ languages, using multilingual embeddings rather than translated keyword overlap.
Deduplication#: Removal of redundant signals — repeated mentions of the same article, syndication copies, retweets — before scoring, so ranking reflects independent observations rather than amplification noise.

Editorial workflow

Editorial brief#: A structured summary of a detected trend prepared for an editor or AI writer. Includes context, cited sources, suggested angle, target audience, and prior coverage by the publisher.
Article generation#: The automated production of a publishable article from an editorial brief, typically a multi-section piece with embedded citations, produced in the order of 60–90 seconds end-to-end.
Multi-CMS publishing#: The capability to publish the same article to multiple content-management systems — WordPress, Ghost, Blogger, Shopify, Wix, Webflow, LinkedIn, Dev.to, Hashnode, Medium — plus Zapier (9,000+ apps) and custom webhooks — through a single connector layer, with per-CMS field mapping.
Multilingual sentiment analysis#: Classification of opinion and emotional tone across 28+ languages without per-language fine-tuning, using a single hybrid model. Used to qualify trends before publication and detect coverage risk.

Discovery & answer engines

llms.txt#: A proposed convention (Answer.AI, 2024) for publishing a markdown index of a site's most important pages, intended for ingestion by large language models. Adoption is partial and not yet confirmed by major LLM editors, but the file is low-effort to ship.
llms-full.txt#: The extractive long-form companion to llms.txt — full plain-text content of key pages flattened into one downloadable file, optimised for LLM training and retrieval.
ai.txt#: A consent-declaration file in the spirit of robots.txt but scoped per usage class (training, inference, search) rather than per crawler. Proposed by Spawning AI.
IndexNow#: An open protocol (Microsoft, Yandex) that lets a website push URL change notifications directly to participating search engines, bypassing crawler discovery. Honored by Bing, Yandex, Seznam; Google does not currently honor IndexNow.
Answer-first content#: A page structure that leads with the direct answer to a question — definitional paragraph, FAQ pair, or summary table — before context and detail. Optimises for extractive citation by answer engines.

Brand & entity signals

Brand entity#: A search engine's structured representation of a brand — distinct from any single page about the brand. The entity graph aggregates name, alternateName, sameAs links, knowsAbout topics, and key relationships. A strong entity wins brand-name SERP queries.
sameAs#: A schema.org Organization property that points at the brand's verified profiles on other platforms — Wikidata, LinkedIn, Crunchbase, X, Product Hunt. Cross-references corroborate the entity in Google Knowledge Graph and answer-engine corpora.
Wikidata#: A free, collaboratively edited knowledge base operated by the Wikimedia Foundation. Used by Google Knowledge Graph and seeded into every major LLM training corpus, making a verified Wikidata entry one of the highest-leverage off-site assets for brand entity recognition.
Knowledge Graph#: Google's structured database of entities — people, places, organizations, concepts — and their relationships. The Knowledge Panel on the right side of a SERP is the public surface of the Knowledge Graph for a given entity.
robots.txt#: The transport-layer file at the site root that declares per-crawler permissions. Distinct from ai.txt (consent declaration per usage class) and llms.txt (content discovery for LLMs).